Example CustomOp2 of torch.searchsorted

[jeromeku]

torch.searchsorted CustomOp2 Implementation and Example

Per #1359, added cpu and cuda implementations of torch.searchsorted as a CustomOp2 as a candle-examples example.

• provides same API as original torch implementation except for sort argument

Important

The tensor being searched MUST be sorted

• includes a kernel_test folder for testing standalone cuda implementation
• also includes jupyter notebook for generating test data from torch.searchsorted

See README.md as well as unittests in searchsorted.rs for documentation and usage examples.